Asrock X570d4u, x570d4u-2L2T discussion thread

I never change a running system, so I never plugged the CPU out. All running fine and well. My RAM was immediately detected and BIOS set 2666 MT/S and I never had any problems. Well, 3200 “overclock” didn’t really work for me but that really was the only hiccup I encountered.

I also never ran the board without all 4 DIMM slots being populated so I don’t know what the temp sensors do when they are empty. I have a cloud running and can’t afford to tinker and change things that often. My baby is in full production and will run until the AIO,PSU or some fans die :slight_smile:

Does anybody a site where ASRock X570D4U-2L2T is still available? Apparently, there is a big shortage of 10Gb NICs and that is why it is out of stock almost everywhere. I only found it on a couple of Dutch sites, but the only country they ship besides Netherlands is Belgium.

Didn’t realize this was happening, but it doesn’t surprise me. It’s kind of buried on their website right now, too. It’s still there, but you pretty much have to know the model number and google it.

In case anyone is interested, here is the information I got through a vendor:

“As informed before, Intel 10G LAN controller is in terribly shortage now, so we can’t offer X570D4U-2L2T now because it’s w/ Intel 10G LAN on board.

We will launch a new motherboard called “X570D4U-2L2T/BCM” which is w/ Broadcom 10G LAN on board.

This new motherboard is still under developing stage, we might have some samples available around October.”

How do Broadcom’s NICs compare to Intel’s NICs?

Intel is usually the go-to if you want compatibility. Broadcom may not have some features, but I’m running a quad Gbit card with broadcom chip and everything is running fine and well in Proxmox, TrueNAS or PfSense. So I didn’t experience lacking Linux and FreeBSD support. 10G chips may be different, so better double check. Otherwise 10G are 10G once everything is running.
Having all 4 NICs from Intel always felt a bit of luxury if you compare it to basically any other board.
If buying today, I would certainly get that /BCM version. Having 10G is more important than having Intel. And there aren’t really alternatives if you want the feature set of this board.

Thank you for your reply. It is good to know that Broadcom NICs play nice with Proxmox since I plan this board for that OS.

I only have experience with 1G NICs. But Broadcom is a large corporation that does mostly server hardware. And most servers run Linux.

But I’m glad to have a Collectors Edition board, now that AsRock is sourcing Broadcom chips for a new board. I bet they didn’t expect those high demands on x470D4U and x570D4U when they were designing them. I’m eager to see a future x670D4U. Great boards.

I got the info from Asrock aswell. Along with a PDF with all revised boards.
New Launch_20220810.zip (2.5 MB)

1 Like

After not using it for a few months, I decided to try ESXi on my x570d4u-2l2t, and I realized that there’s no way to disable loading the video ROM from the PCIe slots like I could with SuperMicro. I had problems trying to use GPU passthrough with ESXi, even with Resizable BAR disabled to avoid guest MMIO space issues.

Honestly, given the lack of Serial-Over-LAN support in the BIOS and IPMI, and the power usage, and the virtual keyboard that sometimes gets ignored, I’m honestly wishing I’d bought something else, maybe a Xeon D.

The IPMI web UI once showed a serial-over-LAN button when I returned to an old tab, only to take it away again when the page finished loading. It seems to be missing the flag that tells the IPMI that it has the SOL feature. And with the beta BIOS, there’s a long delay every boot, while it tries to apply settings from the IPMI’s BIOS settings interface.

I suppose one approach I could take would be to just sell the darn thing…

Hi All

I have the X570D4U MB
Ryzen 9 3900XT
4 x Crucial CT16G4DFD832A - 16GB (1x16GB), PC4-25600 (3200MHz) DDR4, CL22, 1.2v, Single Stick
2 x Samsung 970 Evo Plus m.2
Silverstone SST-RMS03-24 2Ru Case

I have experienced random lockups since building this server last year - however now I need to put this into production (to replace a failing qnap nas).

I’ve tried OVM, TrueNAS Scale, Proxmox 7, Debian and (currently) XCP-NG - and with each OS the machine would randomly lockup after a few days… this time it was 5 days.

I have mostly left the bios settings alone - however, however I do currenly have IMMU enabled for gpu passthrough (just an old 1050Ti). Running the latest non beta bios and bmc firmware. I did try the beta bios this morning… but could not get the machine to boot properly - and it would take forever to boot - eventually causing a kernel panic in xcp-ng - tried resetting the cmos - loading defaults etc… no difference. After wasting 2hrs I went back to the 1.20 firmware.

CPU Temp 41 C (machine is in my garage and it’s winter here).
X570 Temp 57 C

I ran a memtest a while back with no issues.

Any ideas on settings I can try?

I’d consider taking it off your hands. I’m over here in Australia and there’s no stock of this board anywhere and it’s perfect for my needs. Feel free to message me.

I need help configuring this (X570D4U) board’s network. To save energy I’d like to use a single ethernet port but I haven’t succeeded in making that happen. I’m currently using LAN 1 to access Proxmox on the server and LAN 2 to access the BMC. Bonding is disabled.

How do I set this up to use a single NIC and how do I disable the other NICs?

Any help will be appreciated.

It’s not clear from the manual whether LAN1 or LAN2 is optionally shared with the BMC, nor whether it’s fixed or configurable. I think by default—out of the box—if you only plug in one of the two (i.e. leaving LAN_IPMI disconnected) it will work the way you want it to.

Since you already went into the BMC to disable the bond, I’m not sure what to tell you. I would note the IPMI MAC, disconnect LAN_IPMI, then watch on your switch/router for the IPMI MAC. If it doesn’t come back, try LAN2 instead of LAN1. If it still doesn’t come back, you might have to reconnect LAN_IPMI, go back into the BMC, and switch it from eth0 to eth1 or vice-versa and repeat the process.

N.B. that on the 2L2T it’s actually one of LAN3 or LAN4 that’s shared.

Good luck, and please report back!

@bambinone - Thanks for responding. You seem to think that I’m using the IPMI port but I’m not. I’ve made a table that shows the current state although I can’t make any sense out of it. I’ll explain and hope this will help clarify matters.

#1: this is the IP# I’m using to access the server
#2: I can ping 192.168.178.201 but can’t login to server or BMC. Right now I can’t remember where I got the MAC from
#3: 192.168.178.202 is the functioning IPMI IP#
#4: the MAC comes from the eth0 interface in the IPMI network settings panel. I can’t get it to accept an IP# via DHCP
#5: this is the dedicated IPMI port that I’m not using currently

The manual I’m using is located here.

I’ve tried many combinations of connecting the system but this (except for bonding) is the only one that works (I also lost access to the server a few times). I currently have the 2 functioning ports connected to a small switch but that is not optimal as the switch also uses power and I have only a single cable running from the server to the router.

I hope this info will trigger new insights on your end but thanks anyway for making the effort.

In the BMC settings, eth0 is LAN_IPMI and eth1 is LAN1 (I just checked). It looks like you’re already set up and good to go with LAN1 for the BMC, so all you have to do configure the server OS to use the other interface, then you can un-plug LAN2. On the switch/router side you’ll see two different MAC addresses from the same port. The MAC addresses with the d0:f0:99 prefix are for the BMC. (The MAC addresses that I see in my server OS start with a8:a1:59, which doesn’t match any of yours, so not sure what yours are.)

Just for fun I made sure eth1 was enabled in my BMC settings, then I went downstairs and un-plugged LAN_IPMI. LAN2 is passed through to my Forbidden Router and connected directly to my provider’s ONT, so it’s completely out of the picture here. LAN1 is providing both the IPMI connection and the primary interface on the server.

It works great!

juno:~$ ping -c4 tethys
PING tethys (192.168.1.3) 56(84) bytes of data.
64 bytes from tethys (192.168.1.3): icmp_seq=1 ttl=64 time=0.329 ms
64 bytes from tethys (192.168.1.3): icmp_seq=2 ttl=64 time=0.371 ms
64 bytes from tethys (192.168.1.3): icmp_seq=3 ttl=64 time=0.588 ms
64 bytes from tethys (192.168.1.3): icmp_seq=4 ttl=64 time=0.596 ms

--- tethys ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3005ms
rtt min/avg/max/mdev = 0.329/0.471/0.596/0.121 ms

juno:~$ arp -a | grep 192.168.1.3
tethys (192.168.1.3) at a8:a1:59:xx:xx:xx [ether] on br-mgmt

juno:~$ ping -c4 tethys-ipmi
PING tethys-ipmi (192.168.1.102) 56(84) bytes of data.
64 bytes from tethys-ipmi (192.168.1.102): icmp_seq=1 ttl=64 time=0.819 ms
64 bytes from tethys-ipmi (192.168.1.102): icmp_seq=2 ttl=64 time=0.424 ms
64 bytes from tethys-ipmi (192.168.1.102): icmp_seq=3 ttl=64 time=0.567 ms
64 bytes from tethys-ipmi (192.168.1.102): icmp_seq=4 ttl=64 time=0.426 ms

--- tethys-ipmi ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3005ms
rtt min/avg/max/mdev = 0.424/0.559/0.819/0.160 ms

juno:~$ arp -a | grep 192.168.1.102
tethys-ipmi (192.168.1.102) at d0:50:99:e6:9a:5d [ether] on br-mgmt

(DHCP was just used for testing IPMI on this interface, I have eth0 set to a static address.)

I can now see I was approaching this from the wrong direction as I was trying to make the changes through the IPMI whereas if I understand you correctly. I need to make the changes server side.

To be absolutely sure: do I just replace enp38s0 with enp39s0 and enp39s0 with enp38s0 in /etc/network/interfaces and presumably reboot?

auto lo
iface lo inet loopback

iface enp38s0 inet manual

auto vmbr0
iface vmbr0 inet static
        address 192.168.178.200/24
        gateway 192.168.178.1
        bridge_ports enp38s0
        bridge_stp off
        bridge_fd 0

iface enp42s0f3u5u3c2 inet manual

iface enp39s0 inet manual

Well, it’s both. You have to tell the BMC which port(s) you want it to use, and you have to tell the server OS which port(s) you want it to use. And in this case there’s some overlap: only the BMC can see LAN_IPMI, and only the server OS can see LAN2, but both can see LAN1.

Yes, I believe if you just change the bridge definition to use enp39s0, it will work. And that explains why you’re not seeing a MAC address with prefix a8:a1:59 on the other end; you’re seeing the MAC address of the bridge.

Success!

Now using only 1 NIC and I just have to figure out how to switch the other 2 off so I can save some more power. Many thanks!

Was looking at the Fractal Node 804 for an ASRock Rack X570 mobo with Ryzen 3500x. I really appreciate all the comments in this thread!

The plan is to upgrade my TrueNAS Scale server with this mobo and 3500x.

Question for folks: is it possible to disable BMC and use the ASpeed chip as a GPU? The 3500X has no iGPU, and I was hoping to ditch the discrete graphics card and go headless for power savings.

After all the issues people have posted about the BMC, I would like to stick with my PiKVM :slight_smile:

I’m trying to reduce power for my 3700X server on a X570D4U board. As I understand it, higher C-states should make the system more efficient. However, if I run powertop, I never see C-states higher than 2. The OS is Linux 5.15.53-1-pve #1 SMP PVE 5.15.53-1. Global C-state Control is set to AUTO (default) but enabling does not improve anything (perhaps the opposite).

Can someone with a similar system run powertop and share the result?


           Pkg(OS)  |            Core(OS) |            CPU(OS) 0   CPU(OS) 8
POLL        0.0%    | POLL        0.0%    | POLL        0.0%    0.0 ms  0.0%    0.0 ms
C1          0.2%    | C1          0.2%    | C1          0.3%    0.5 ms  0.1%    0.2 ms
C2         98.2%    | C2         99.1%    | C2         98.8%   18.8 ms 99.4%   16.5 ms

                    |            Core(OS) |            CPU(OS) 1   CPU(OS) 9
                    | POLL        0.0%    | POLL        0.0%    0.0 ms  0.0%    0.0 ms
                    | C1          0.3%    | C1          0.6%    0.2 ms  0.0%    0.1 ms
                    | C2         96.7%    | C2         95.9%    5.1 ms 97.5%   20.2 ms

                    |            Core(OS) |            CPU(OS) 2   CPU(OS) 10
                    | POLL        0.0%    | POLL        0.0%    0.0 ms  0.0%    0.0 ms
                    | C1          0.2%    | C1          0.1%    0.1 ms  0.4%    0.6 ms
                    | C2         98.6%    | C2         98.9%   13.3 ms 98.4%   26.0 ms

                    |            Core(OS) |            CPU(OS) 3   CPU(OS) 11
                    | POLL        0.0%    | POLL        0.0%    0.0 ms  0.0%    0.0 ms
                    | C1          0.3%    | C1          0.2%    0.6 ms  0.4%    1.0 ms
                    | C2         97.5%    | C2         98.8%   23.8 ms 96.3%   19.9 ms

                    |            Core(OS) |            CPU(OS) 4   CPU(OS) 12
                    | POLL        0.0%    | POLL        0.0%    0.0 ms  0.0%    0.0 ms
                    | C1          0.2%    | C1          0.3%    0.3 ms  0.0%    0.3 ms
                    | C2         97.1%    | C2         98.5%   21.3 ms 95.7%   45.5 ms

                    |            Core(OS) |            CPU(OS) 5   CPU(OS) 13
                    | POLL        0.0%    | POLL        0.0%    0.0 ms  0.0%    0.0 ms
                    | C1          0.1%    | C1          0.1%    0.2 ms  0.1%    0.2 ms
                    | C2         99.4%    | C2         99.3%   14.6 ms 99.6%   24.8 ms

                    |            Core(OS) |            CPU(OS) 6   CPU(OS) 14
                    | POLL        0.0%    | POLL        0.0%    0.0 ms  0.0%    0.0 ms
                    | C1          0.2%    | C1          0.1%    0.5 ms  0.3%    0.5 ms
                    | C2         98.7%    | C2         99.0%   15.5 ms 98.4%   23.5 ms

                    |            Core(OS) |            CPU(OS) 7   CPU(OS) 15
                    | POLL        0.0%    | POLL        0.0%    0.0 ms  0.0%    0.0 ms
                    | C1          0.3%    | C1          0.2%    0.3 ms  0.4%    0.5 ms
                    | C2         97.9%    | C2         96.7%    7.7 ms 99.2%   21.4 ms